Introduces average rate of change as the slope of a secant line, computed from a table, a graph or a formula, with units read off the fraction. Builds the difference quotient as the same slope with the gap named h, then supplies the vocabulary for describing a graph precisely: increasing and decreasing intervals, local extrema, and absolute extrema, including why the endpoints of a closed window matter.
Subject: Precalculus · 69 slides · symbolic lesson
Open the interactive version of this deck
Title
Precalculus · Section 1.3
Turning a picture of a function into precise, checkable sentences
Objectives
Sections 1.1 and 1.2 asked what a function is. This one asks how a function behaves: how fast it changes, where it rises, where it falls, and where it is at its best and worst.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 66-79 — learning objectives 1.3.1 to 1.3.4
Everything here is about describing a function precisely once you have its graph, its table or its rule. That sounds like a soft skill, and it is not: almost every exam question on this section is marked on the exact wording, not on the arithmetic.
Two habits will carry you through it. The first is to say out loud whether a quantity is an input or an output before you write it down. Intervals are made of inputs; extreme values are outputs; getting those the wrong way round is the single most common lost mark in this section. The second is to attach units to every rate you compute, because the units are read straight off the fraction and they tell you instantly whether you divided the right way up.
The last two parts, on local and absolute extrema, are also where calculus is heading. You will find turning points by reading a graph here; in Chapter 12 and in Calculus I you will find them exactly with algebra. The vocabulary you build now is the vocabulary you will use then.
Warm-up
A car covers 120 miles in 2 hours. Almost everyone can produce the number 60. The question is what that number is actually claiming.
Discussion prompt
Does saying the car averaged 60 miles per hour mean it was ever travelling at exactly 60 miles per hour? Write yes or no, and one sentence of reason.
Write your answer before revealing. Most people say yes without thinking, and the reason it feels obviously true is worth examining.
The honest answer is no. The car covered 120 miles in 2 hours, and that is all the two numbers tell you. It could have crawled for an hour and then raced for an hour; it could have stopped for lunch and made the time up. An average rate of change is computed from the two ends of an interval, and it knows nothing whatever about the middle.
Hold on to that, because the rest of the lesson is built around it. The first part of the lesson computes these blind averages; the third and fourth parts supply the language for describing what actually happened in between.
Section
Part 1
Concept
A rate of change says how much the output moves for each unit the input moves. To get one from two data points, divide the change in output by the change in input.
\[ \text{average rate of change} = \frac{\text{change in output}}{\text{change in input}} = \frac{\Delta y}{\Delta x} = \frac{f(x_2) - f(x_1)}{x_2 - x_1} \]
Figure (svg): Eight dots showing the average price of a gallon of gasoline each year from 2005 to 2012, zig-zagging between 2.31 and 3.68 dollars, with a straight dashed line joining the first dot to the last.
The Greek letter delta is read as the change in. Over the whole stretch from 2005 to 2012 the price rose by 1.37 dollars in 7 years, so the average rate of change was about 19.6 cents per year.
\[ \frac{\Delta y}{\Delta x} = \frac{1.37}{7} \approx 0.196 \text{ dollars per year} \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 66-67 — Table 1.17 and equation 1.1
The formula itself is not new. It is the slope formula you learned for straight lines, applied to two points that happen to sit on a curve. What is new is the interpretation: it is a rate, and a rate has units.
Look at the dots in the figure. The price of a gallon of gasoline jumped about between 2005 and 2012: up, up, up, sharply down in 2009, then up again. The dashed line is the only thing the average rate of change over that whole stretch sees. It reports a steady rise of about 19.6 cents a year, which is true as a summary and false as a description of any single year.
About the notation: the Greek letter delta is a shorthand for the change in, and delta y is one symbol standing for a single number, the difference of two outputs. It is not delta multiplied by y. You will sometimes see delta f written instead of delta y, which means exactly the same thing.
Notation
Annotate
On: \( \frac{\Delta y}{\Delta x} = \frac{f(x_2) - f(x_1)}{x_2 - x_1} \)
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 67 — Rate of Change box, equation 1.1
Work through the four annotations in order; each one is a place where marks get lost. The first two are about which quantity goes where, and the rule never changes: outputs on top, inputs underneath.
The third annotation is the one that bites. Both subtractions must be taken in the same order. If you decide that the 2007 data is point one and the 2009 data is point two, then both the numerator and the denominator must be second minus first. Reversing both is harmless, because the two sign changes cancel; reversing only one flips the answer's sign.
The fourth is a free self-check you should use every single time. Say the units out loud: dollars per year, miles per hour, newtons per centimetre. If what comes out of your mouth is years per dollar, you have the fraction upside down and you know it before you write anything else.
Picture it
Figure (svg): A curve rising, dipping and rising again, with two marked points at x equals 1 and x equals 5. A straight line joins them, and below it a dashed horizontal segment of length 4 and a dashed vertical segment of height 3 form a right triangle.
Mark the two points on the graph and join them with a straight line. That line is called a secant line, and the fraction you just computed is exactly its slope: rise over run, the slope formula from Algebra 1 applied to two points of a curve.
\[ \text{slope of the secant through } (x_1, f(x_1)) \text{ and } (x_2, f(x_2)) = \frac{\Delta y}{\Delta x} \]
This is the picture to carry for the rest of the lesson. Put a dot on the graph at each of the two inputs and draw the straight line through them. That line is a secant line, and the number you just computed is its slope.
The dashed triangle under the line is the whole formula, drawn. The horizontal leg is the change in input, four units here, and the vertical leg is the change in output, three units, so the slope is three quarters. Rise over run and change in output over change in input are the same instruction in two vocabularies.
Notice how far the curve wanders from the line between the two points: it rises, dips, and rises again while the line just climbs steadily. That wandering is exactly what the word average is apologising for.
Worked example
Using the prices in Table 1.17, find the average rate of change of the price of gasoline between 2007 and 2009.
| year | 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 |
|---|---|---|---|---|---|---|---|---|
| C(y) | 2.31 | 2.62 | 2.84 | 3.30 | 2.41 | 2.84 | 3.58 | 3.68 |
Read the two outputs
Why: The price in 2007 and the price in 2009; nothing in between is used.
\[ y_1 = 2007,\; C(y_1) = 2.84, \qquad y_2 = 2009,\; C(y_2) = 2.41 \]
Subtract the outputs
Why: Second minus first, so a drop comes out negative.
\[ \Delta y = 2.41 - 2.84 = -0.43 \]
Subtract the inputs in the same order
Why: Second year minus first year.
\[ \Delta x = 2009 - 2007 = 2 \]
Figure (svg): A close-up of the gasoline prices for 2007 to 2009: 2.84, then 3.30, then 2.41 dollars, joined by a straight line that falls from the first to the last, with the drop of 0.43 dollars marked over the two-year run.
Divide and attach the units
Why: Dollars on top, years underneath.
\[ \frac{\Delta y}{\Delta x} = \frac{-0.43}{2} = -0.215 \text{ dollars per year} \]
Check the sign against the picture
Why: The secant falls from left to right, so the rate must be negative, and it is. The book rounds this to minus 0.22 dollars per year; both are the same number to the nearest cent.
\[ 2.41 < 2.84 \;\Longrightarrow\; \Delta y < 0 \;\Longrightarrow\; \frac{\Delta y}{\Delta x} < 0 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 67 — Example 1.29
Two things make this example worth doing carefully. The first is that only two of the eight numbers in the table are used. The 2008 price of 3.30 is the highest in the whole stretch, and it plays no part at all; look at the figure and see how far above the line that dot sits.
The second is the sign. Prices fell over these two years, so the change in output is negative, and a negative divided by a positive is negative. A negative rate of change is a perfectly ordinary answer: it means the output falls as the input rises. It does not mean you made a mistake.
The answer comes out as minus 0.215 dollars per year exactly. The book rounds it to minus 0.22 dollars per year, which is the same number to the nearest cent; either is fine as long as you say which you have done. In words: over those two years the price dropped by about 22 cents a year on average.
Prediction
\[ C(2005) = 2.31, \qquad C(2010) = 2.84 \]
Predict first
What is the average rate of change of the price between 2005 and 2010?
Correct: About 0.106 dollars per year
Why: The price rose by 0.53 dollars over 5 years, and 0.53 divided by 5 is 0.106 dollars per year. The change itself, 0.53, is a common wrong answer because the division by the run gets skipped.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 67 — Try It 1.21
Commit to one option before revealing. The two prices you need are 2.31 and 2.84, and the interval is five years long, not seven.
The change in output is 0.53 dollars. The tempting wrong answer is to stop right there and report 0.53, because it feels like the answer to how much the price changed. It is, but the question asked for a rate, which means the change per year, so it still has to be divided by the run of 5.
The result, about 0.106 dollars per year, is roughly half the 0.196 you got over 2005 to 2012. Same starting year, different finishing year, different answer. Every average rate of change belongs to the particular interval it was computed on, and quoting one without its interval is meaningless.
Intuition
Figure (svg): Two different curves drawn between the same two points, at 1 and at 5. One climbs early and flattens; the other dips below and then climbs steeply. The straight line joining the shared endpoints is drawn through both.
Two functions can agree at the two endpoints and do completely different things in between, and their average rates of change over that interval will be identical. The average is a summary, and like every summary it throws information away.
This is why the word average is in the name, and why an average rate of change is always quoted together with the interval it was computed on. Change the interval and you generally change the number.
Look hard at the two curves. They start at the same point and finish at the same point, and in between one of them shoots up and flattens while the other sags below the start line before racing up at the end. Their average rates of change over this interval are identical.
This is not a defect to be fixed; it is what an average is for. When you are told a car averaged 60 miles an hour over a trip, you are being given a summary, and a summary is useful precisely because it is short. The price is that it cannot be reversed: from the number alone you cannot rebuild the journey.
Two practical consequences. First, always quote the interval alongside the rate. Second, if someone asks what the function was doing rather than where it ended up, an average rate of change is the wrong tool, and the vocabulary in Part 3 is the right one.
Worked example
The graph shows a function g. Find the average rate of change on the interval from minus 1 to 2.
Figure (svg): The parabola g of t equals t minus one, squared, with marked points at t equals minus one, height 4, and t equals 2, height 1. A straight line joins them, falling; a dashed horizontal run of 3 and a dashed vertical drop of 3 are drawn.
Read the output at the left endpoint
Why: Go up from t equal to minus 1 until you hit the curve, then across to the vertical axis.
\[ g(-1) = 4 \]
Read the output at the right endpoint
Why: Same procedure at t equal to 2.
\[ g(2) = 1 \]
Form the fraction
Why: Outputs on top, inputs underneath, both second minus first.
\[ \frac{\Delta g}{\Delta t} = \frac{1 - 4}{2 - (-1)} \]
Simplify
Why: Minus three over three.
\[ \frac{-3}{3} = -1 \]
Check against the slope triangle
Why: The dashed run is 3 units right and the dashed rise is 3 units down, so the secant drops one unit for every unit it moves right: slope minus 1, matching the arithmetic.
\[ \text{run } 3, \; \text{fall } 3 \;\Longrightarrow\; \text{slope} = -1 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 68 — Example 1.30 and Figures 1.43, 1.44
The arithmetic here is trivial; the skill is reading the two heights off the grid correctly. Go up from the input on the horizontal axis until you meet the curve, then straight across to the vertical axis. At the input minus 1 that gives 4, and at the input 2 it gives 1.
Now watch the denominator. The run is 2 minus negative 1, which is 3, not 1. Subtracting a negative number is where this example catches people, so write the brackets in: two minus open bracket minus one close bracket.
The slope triangle in the figure is your check. The run is three units to the right and the fall is three units down, so the line drops one unit for every unit it moves right, and the slope is minus 1. Whenever you compute a rate from a graph, glance at the line: falling means negative, rising means positive, and that catches a sign error in a second.
Trap
A common wrong line, using Example 1.30:
\[ \frac{g(2) - g(-1)}{-1 - 2} \]
\[ = \frac{-3}{-3} = 1 \]
Wrong. The outputs were subtracted second-minus-first, the inputs first-minus-second.
Fix the two points as point one and point two and keep them fixed in both subtractions. Reversing both is harmless, because the two sign changes cancel.
\[ \frac{g(-1) - g(2)}{-1 - 2} = \frac{3}{-3} = -1 \]
The picture is the referee: a falling secant cannot have a positive slope.
This is the most common arithmetic mistake in the whole section, and it is invisible in the answer: you get a perfectly reasonable-looking number with the wrong sign, and nothing in the algebra complains.
The cure is procedural. Before you compute anything, label the two points: this one is point one, that one is point two. Then both differences are second minus first, with no further thought. If you would rather subtract the other way round, do it in both places; reversing both is completely safe because the two sign changes cancel.
The picture is the referee. A secant that falls from left to right has a negative slope, full stop. If your arithmetic hands you a positive number for a falling line, you have swapped an order somewhere.
Worked example
After picking up a friend who lives 10 miles away, Anna records her distance from home over time. Find her average speed over the first 6 hours.
| t (hours) | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|
| D(t) (miles) | 10 | 55 | 90 | 153 | 214 | 240 | 292 | 300 |
Identify the interval and its two outputs
Why: The first 6 hours means from t equal to 0 to t equal to 6.
\[ D(0) = 10, \qquad D(6) = 292 \]
Subtract the outputs
Why: She started 10 miles from home, so the distance covered is not 292.
\[ \Delta D = 292 - 10 = 282 \text{ miles} \]
Divide by the elapsed time
Why: Six hours of driving.
\[ \frac{282}{6 - 0} = 47 \text{ miles per hour} \]
Figure (svg): Eight dots showing distance from home in miles at each hour from 0 to 7, rising from 10 to 300. A dashed line joins the first dot to the dot at hour 6, and a short steeper solid segment joins the dots at hours 2 and 3.
Check on a shorter interval
Why: Between hours 2 and 3 she covered 153 minus 90, that is 63 miles in one hour, faster than her 47 mile per hour average, which matches the steeper red segment. Note the printed Table 1.18 shows 282 at t equal to 6, but the printed solution computes 292 minus 10 and states 282 miles travelled: 292 is the intended table entry, and 47 is the correct answer.
\[ \frac{D(3) - D(2)}{3 - 2} = \frac{153 - 90}{1} = 63 > 47 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 69 — Example 1.31 and Table 1.18
Read the question carefully: her average speed over the first six hours. Speed is distance travelled per hour, and she did not start at home; she started ten miles away. So the distance covered is the difference of the two readings, not the reading at hour six.
That distinction is the entire point of the example. The output is her distance from home, and the change in output is 282 miles, over a change in input of 6 hours. Dividing gives 47 miles per hour.
The check is worth copying. Take a shorter interval, hours 2 to 3, and the average speed over that single hour is 63 miles per hour, well above her overall average. That matches the short steep segment in the figure and tells you something the overall average hid: she was moving faster in the middle of the trip than at the start. One warning about the book here. Its Table 1.18 prints 282 at hour six, but its own solution computes 292 minus 10 and says she travelled 282 miles, so 292 is the intended table entry and 47 is the intended answer. If you use the printed 282 you get 45.33 instead.
Concept
Every rate of change carries units, and they are read straight off the fraction: the output's units divided by the input's units. Saying them out loud catches an upside-down fraction instantly.
| situation | output per input | the rate |
|---|---|---|
| a population of rats grows | rats per week | 40 rats per week |
| a car travels | miles per hour | 68 miles per hour |
| a car burns fuel | miles per gallon | 27 miles per gallon |
| voltage drives current | amperes per volt | 0.125 amperes per volt |
| a college account is spent | dollars per quarter | a fall of 4000 dollars per quarter |
A negative rate is a genuine rate, not an error. It means the output falls as the input rises.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 67 — other examples of rates of change
Units are not decoration. They are a free, instant check on whether the fraction is the right way up, and on some exams they carry their own mark.
The rule is mechanical: take the units of the output, then the words per, then the units of the input. Rats per week, miles per hour, miles per gallon, amperes per volt, dollars per quarter. In every case the thing after the word per is the input.
Notice the last row of the table. Money leaving an account is a negative rate, and it is still a rate. Negative simply means that as the input grows the output shrinks. Nothing in the definition prefers positive answers.
Sorting
Sort into buckets
In each pair the input is named. Sort each quantity into the numerator or the denominator of the rate.
Sort all six before checking. Three of them are outputs and three are inputs, and the pairs are deliberately made of the same two quantities in different roles.
The fastest way through is the word per. If you are asked for miles per hour, then miles are on top and hours underneath; the input is always the quantity named after per. Items c and d are the same pair from a shop's point of view: cost is what the shop charges once you have chosen the number of items, so cost is the output and the number of items is the input.
Items b, d and f may look like trick questions, and in a sense they are: they name the input itself. Whatever is declared to be the input goes in the denominator, even when that feels circular. Keeping that straight is what stops you writing hours per mile.
Worked example
Compute the average rate of change of the function below on the interval from 2 to 4.
\[ f(x) = x^2 - \frac{1}{x} \]
Evaluate at the left endpoint
Why: Substitute 2 and combine over a common denominator.
\[ f(2) = 2^2 - \frac12 = 4 - \frac12 = \frac72 \]
Evaluate at the right endpoint
Why: Substitute 4 the same way.
\[ f(4) = 4^2 - \frac14 = 16 - \frac14 = \frac{63}{4} \]
Subtract the outputs
Why: Common denominator 4: seven halves is fourteen quarters.
\[ \frac{63}{4} - \frac{7}{2} = \frac{63 - 14}{4} = \frac{49}{4} \]
Divide by the change in input
Why: The run is 4 minus 2, that is 2; dividing by 2 halves the fraction.
\[ \frac{\frac{49}{4}}{2} = \frac{49}{8} \]
Figure (svg): The curve y equals x squared minus one over x, rising from about 3.5 at x equals 2 to about 15.75 at x equals 4, with the two points marked and a straight line joining them.
Check numerically
Why: Forty-nine eighths is 6.125, and the curve does climb from 3.5 to 15.75, a rise of 12.25 over a run of 2, which is 6.125.
\[ \frac{15.75 - 3.5}{2} = \frac{12.25}{2} = 6.125 = \frac{49}{8} \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 70 — Example 1.32
When the function is given by a rule rather than a table, there is one extra step at the start: you have to compute the two outputs yourself. Everything after that is identical.
Resist the urge to switch to decimals. Seven halves and sixty-three quarters are exact, and keeping them as fractions makes the subtraction clean: put both over 4 and the numerator becomes 49 over 4. Then dividing by the run of 2 just doubles the denominator, giving 49 over 8.
The check converts to decimals only at the end, which is the right moment. The heights are 3.5 and 15.75, a rise of 12.25 over a run of 2, and 6.125 is exactly 49 over 8. If your fraction work and your decimal work disagree, the fraction work is usually where the slip is, and it is usually a common denominator.
Fill the middle
\[ f(x) = x - 2\sqrt{x}, \qquad \text{on the interval from } 1 \text{ to } 9 \]
Fill in the blanks
\frac31/2 = \frac___}) - (-1)}___ = ___
Why: The square root of 9 is 3, so f(9) is 9 minus 6, which is 3. At the other end the square root of 1 is 1, so f(1) is 1 minus 2, which is minus 1. The rise is 3 minus minus 1, that is 4, over a run of 8, giving one half.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 70 — Try It 1.22
Fill both blanks before checking. The only real work is evaluating the rule at the two endpoints, and the square roots are both whole numbers here, which is why these endpoints were chosen.
At the input 9 the square root is 3, so twice it is 6, and 9 minus 6 is 3. At the input 1 the square root is 1, so twice it is 2, and 1 minus 2 is minus 1. Watch the double negative in the numerator: 3 minus negative 1 is 4, not 2.
Four over a run of 8 is one half. Notice that the function is negative at one endpoint and positive at the other, and the rate still comes out positive, because all that matters is that the output ended higher than it started.
Worked example
The electrostatic force in newtons between two charged particles a distance d centimetres apart is given below. Find the average rate of change of the force as the distance grows from 2 centimetres to 6 centimetres.
\[ F(d) = \frac{2}{d^2} \]
Evaluate at both endpoints
Why: Square the distance first, then divide.
\[ F(2) = \frac{2}{2^2} = \frac12, \qquad F(6) = \frac{2}{6^2} = \frac{2}{36} = \frac{1}{18} \]
Subtract the outputs
Why: Common denominator 36: one half is eighteen thirty-sixths.
\[ \frac{2}{36} - \frac{18}{36} = -\frac{16}{36} = -\frac49 \]
Divide by the change in distance
Why: The run is 6 minus 2, that is 4.
\[ \frac{-\frac49}{4} = -\frac{4}{36} = -\frac19 \]
Figure (svg): The curve F equals two over d squared falling steeply from 0.5 newtons at 2 centimetres to about 0.06 newtons at 6 centimetres, with the two points marked and a shallow straight line joining them.
Check the size and the units
Why: One ninth of a newton per centimetre is about 0.111, and the curve really does drop about 0.444 newtons over 4 centimetres. Force divided by distance gives newtons per centimetre, as it must.
\[ \frac{0.0556 - 0.5}{4} = -0.1111 \approx -\frac19 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 70 — Example 1.33
This is the first example where the units carry real meaning. Force in newtons is the output and distance in centimetres is the input, so the rate is in newtons per centimetre: how much force is lost for each extra centimetre of separation.
The fraction work is the same pattern as the last example. Square first, then divide: at 2 centimetres the force is one half a newton, and at 6 centimetres it is two thirty-sixths, which is one eighteenth. Put both over 36, subtract to get minus 16 over 36, then divide by the run of 4.
The answer, minus one ninth of a newton per centimetre, is negative because the force falls off as the particles move apart. Look at the shape of the curve, though: almost all of that loss happens in the first centimetre or two. The average over the whole four centimetres badly understates how steeply the force drops near the start, which is once again the average hiding the middle.
Estimation
Figure (svg): A curve leaving the height 3 at the input 0 almost flat, then bending sharply upward to the height 23 at the input 4, with a straight line joining those two points.
Predict first
The curve leaves the height 3 at the input 0 and reaches the height 23 at the input 4. Roughly what is its average rate of change over that interval?
Correct: About 5
Why: The output climbed by about 20 while the input climbed by 4, so the rate is near 5 output units per input unit. Twenty is the rise on its own, with the division forgotten, and 0.2 is the fraction turned upside down.
Pick an answer before revealing. Estimating first is a habit worth building, because it costs about three seconds and it catches the two errors that actually happen.
Here the output climbed by about 20 while the input climbed by 4, so the answer has to be near 5. The option 20 is the rise on its own, which is what you get if you forget to divide by the run; the option 0.2 is the fraction upside down. Both of those are caught by a glance at the picture, and neither is caught by re-checking your arithmetic.
Notice how far the curve sits below its own secant line here. Over most of this interval the function is changing far more slowly than 5 units per unit, and then far more quickly at the end. The average of 5 is a fair summary and a poor description.
Worked example
Using the graph, estimate the average rate of change from the input 1 to the input 4, and then from 2 to 5.
Figure (svg): A wavy curve on a grid from 0 to 8 passing through heights 2, 5, 3, 1, 4, 7, 5, 3 and 5 at the whole numbers. Two straight lines are drawn: one from the point at 1 down to the point at 4, and one from the point at 2 up to the point at 5.
Read the four heights off the grid
Why: Each of the four points sits on a grid intersection.
\[ f(1) = 5, \; f(4) = 4, \qquad f(2) = 3, \; f(5) = 7 \]
Do the first interval
Why: From 1 to 4 the output falls by 1 over a run of 3.
\[ \frac{f(4) - f(1)}{4 - 1} = \frac{4 - 5}{3} = -\frac13 \]
Do the second interval
Why: From 2 to 5 the output rises by 4 over a run of 3.
\[ \frac{f(5) - f(2)}{5 - 2} = \frac{7 - 3}{3} = \frac43 \]
Check the signs against the two lines
Why: The red line, from 1 to 4, tilts gently downward, so a small negative number is right. The green line, from 2 to 5, climbs steeply, so a number bigger than 1 is right. Same graph, same length of interval, completely different rates.
\[ -\frac13 < 0 < \frac43 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 79-80 — Exercises 169 and 170, Figure 1.57
Every value you need sits exactly on a grid intersection, which is how these exercises are always built. Find the input on the horizontal axis, go straight up to the curve, then straight across.
The two intervals here are both three units long, which makes the comparison clean. From 1 to 4 the output slips from 5 down to 4, so the rate is minus one third: a gentle decline. From 2 to 5 the output climbs from 3 to 7, so the rate is four thirds: a brisk rise, four times as steep and in the opposite direction.
Two intervals of the same length on the same graph, giving completely different rates. If you ever catch yourself talking about the average rate of change of a function, with no interval named, this figure is the reason that phrase means nothing.
Section
Part 2
Concept
So far the two inputs have been two numbers. Now fix the first input, call it a, and describe the second one as a plus h, where h is the gap. The average rate of change becomes a formula rather than a number.
\[ \frac{f(a + h) - f(a)}{(a + h) - a} = \frac{f(a + h) - f(a)}{h}, \qquad h \ne 0 \]
difference quotient — The average rate of change of f from a to a plus h: the slope of the secant through the points of the graph at a and at a plus h.
Figure (svg): The parabola y equals x squared with a fixed point at x equals 1. Three straight lines leave that point and reach the curve at x equals 3, x equals 2 and x equals 1.5, each less steep than the one before.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 70-71 — Examples 1.34 onward; exercises 158 to 168
Everything in this part is the same calculation you have been doing, with one change of notation: instead of naming both inputs as numbers, you name the first one a and describe the second as a plus h, where h is the gap between them.
Do the denominator carefully once and you never have to do it again. The change in input is a plus h, minus a, and that is simply h. So the run of every difference quotient is h, which is why h is the only thing underneath.
In the figure the left endpoint is nailed down at the input 1 and the right one sits at distances 2, 1 and 0.5 away. Three different gaps, three different secant lines, three different slopes, all from the same fixed point. Because the answer now depends on h, it is a formula rather than a number, and that is exactly what makes it useful.
Intuition
Two reasons, one immediate and one for later. The immediate one is that a formula answers infinitely many questions at once: once you know the rate from a to a plus h in general, every particular interval is a substitution away.
The later reason is the whole of calculus. As the gap shrinks, the second point slides back along the curve towards the first and the secant pivots towards the tangent. The number it pivots towards is the instantaneous rate of change, the derivative. This lesson stops one step short of that limit, but it builds the expression the limit is taken of.
So the algebra below is not busywork. Every one of these problems ends with the letter h still present and every term carrying a factor of h that has already cancelled once. That cancellation is the point.
There is a short-term payoff and a long-term one. The short-term payoff is that one formula answers every question about that function at once: work out the difference quotient of the squaring rule as 2x plus h, and you can read off the rate over any interval you like by substituting.
The long-term payoff is calculus. Watch the three secants in the previous figure: as the gap shrinks, the second point slides back down the curve towards the first, and the line pivots. The slopes were 4, then 3, then 2.5, and they are clearly heading somewhere. Where they head is the slope of the tangent line at that point, which is the instantaneous rate of change.
This lesson stops one step short of taking that limit, but it builds the expression the limit will be taken of. That is why every one of these problems is set up so that the h in the denominator cancels: without the cancellation you could not let h become zero, and with it you can.
Worked example
Find the average rate of change of the function below on the interval from 0 to a. The answer will be an expression involving a.
\[ g(t) = t^2 + 3t + 1 \]
Write the formula with the two endpoints
Why: The run is a minus 0, that is a.
\[ \frac{g(a) - g(0)}{a - 0} \]
Evaluate both outputs
Why: Substitute a, then substitute 0.
\[ g(a) = a^2 + 3a + 1, \qquad g(0) = 0^2 + 3(0) + 1 = 1 \]
Subtract and simplify the numerator
Why: The two constant ones cancel.
\[ \frac{(a^2 + 3a + 1) - 1}{a} = \frac{a^2 + 3a}{a} \]
Factor and cancel
Why: Both terms on top carry a factor of a.
\[ \frac{a(a + 3)}{a} = a + 3 \]
Check on a particular interval
Why: On the interval from 0 to 5 the formula predicts 8. Directly: g of 5 is 41 and g of 0 is 1, so the rate is 40 over 5, which is 8.
\[ \frac{g(5) - g(0)}{5} = \frac{41 - 1}{5} = 8 = 5 + 3 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 70-71 — Example 1.34
This is the gentlest possible introduction to an answer that is an expression rather than a number, because one of the endpoints is still a plain number. The input 0 makes the run just a, and it makes the output easy: substituting 0 into the rule leaves only the constant term, 1.
Watch the cancellation. The constant 1 appears in both outputs, so it disappears in the subtraction; then every surviving term in the numerator carries a factor of a, so factoring lets the a underneath cancel. That two-stage pattern, constants cancel then a common factor cancels, is what happens in every one of these problems.
The check is the habit to build. You have a formula claiming to give the rate on any interval from 0 to a; test it on one you can compute directly. On the interval from 0 to 5 the formula predicts 8, and doing it the long way gives 41 minus 1 over 5, which is also 8. That takes ten seconds and catches almost any algebra slip.
Worked example
Find the average rate of change of the function below on the interval from 5 to a.
\[ f(x) = x^2 + 2x - 8 \]
Evaluate at the fixed endpoint
Why: Substitute 5: twenty-five plus ten minus eight.
\[ f(5) = 25 + 10 - 8 = 27 \]
Write the quotient
Why: Outputs on top, inputs underneath.
\[ \frac{(a^2 + 2a - 8) - 27}{a - 5} = \frac{a^2 + 2a - 35}{a - 5} \]
Factor the numerator
Why: Two numbers multiplying to minus 35 and adding to 2: seven and minus five.
\[ a^2 + 2a - 35 = (a + 7)(a - 5) \]
Cancel the common factor
Why: The factor a minus 5 is exactly the denominator.
\[ \frac{(a + 7)(a - 5)}{a - 5} = a + 7 \]
Check on a particular interval
Why: Take a equal to 9. The formula gives 16. Directly: f of 9 is 91, f of 5 is 27, and 64 over 4 is 16.
\[ \frac{f(9) - f(5)}{9 - 5} = \frac{91 - 27}{4} = 16 = 9 + 7 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 71 — Try It 1.23
Same idea, but now the fixed endpoint is 5 rather than 0, which means the denominator is a minus 5 and the constants no longer cancel on their own. Evaluate the rule at 5 first and get a single number, 27, out of the way.
The numerator is a quadratic in a, and it must factor with a minus 5 as one of its factors, because the quotient has to simplify. That is a useful piece of foresight: if your numerator does not have the denominator as a factor, look again at your arithmetic before you look for a clever trick.
Notice the answer, a plus 7. Compare it with the previous example's a plus 3, and with the fact that the fixed endpoints were 5 and 0 while the rules were x squared plus 2x minus 8 and t squared plus 3t plus 1. In both cases the answer is the sum of the two inputs plus the coefficient of the linear term. That pattern is not a coincidence, and the difference quotient of the squaring rule on the next few slides explains it.
Notation
Annotate
On: \( \frac{f(x + h) - f(x)}{h} \)
Every difference quotient of a polynomial ends with a factor of h cancelling. If nothing cancels, look for an algebra slip in the expansion.
The first annotation is the one that costs the most marks in this whole lesson. Substituting x plus h into a rule means replacing every single x by the whole expression x plus h, kept in brackets until you expand it. It does not mean taking f of x and adding h.
The third annotation is worth a second look too. Once you have written the quotient, the denominator is just h. You never carry x plus h minus x around; you simplify it at the start and forget it.
The last annotation quietly explains the whole design of the section. The gap must not be zero, because you cannot divide by zero, but it may be as small as you like. Everything is arranged so that after simplification the h in the denominator has gone, and only then does it make sense to ask what happens as the gap closes.
Worked example
Find the average rate of change on the interval from x to x plus h.
\[ f(x) = 2x^2 + 1 \]
Substitute the shifted input
Why: Brackets around the whole expression before squaring.
\[ f(x + h) = 2(x + h)^2 + 1 \]
Expand the square
Why: The square of a binomial has three terms.
\[ 2(x^2 + 2xh + h^2) + 1 = 2x^2 + 4xh + 2h^2 + 1 \]
Subtract the original output
Why: The two x squared terms and the two ones both cancel.
\[ (2x^2 + 4xh + 2h^2 + 1) - (2x^2 + 1) = 4xh + 2h^2 \]
Divide by h and cancel
Why: Both surviving terms carry a factor of h.
\[ \frac{4xh + 2h^2}{h} = 4x + 2h \]
Check with numbers
Why: Take x equal to 3 and h equal to 0.1. The formula gives 12.2. Directly, f of 3.1 is 20.22 and f of 3 is 19, and 1.22 over 0.1 is 12.2.
\[ \frac{f(3.1) - f(3)}{0.1} = \frac{20.22 - 19}{0.1} = 12.2 = 4(3) + 2(0.1) \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 80 — Exercise 162
This is the model problem for the whole part; every other quadratic works exactly the same way. Substitute, expand, subtract, cancel.
The expansion is where care is needed. The square of x plus h is x squared plus 2xh plus h squared, three terms, and the middle one is the one that survives to become the answer. Doubling all of that and adding 1 gives the shifted output.
Then watch what cancels. The 2x squared appears in both outputs and goes; the plus 1 appears in both and goes. What is left, 4xh plus 2h squared, has a factor of h in every term, so dividing by h leaves 4x plus 2h with no fraction at all. The check with x equal to 3 and h equal to 0.1 takes a moment and confirms all of it.
Faded example
Now the same rule with a linear term attached. Two lines of the solution are blank.
\[ f(x) = 2x^2 - 3x \]
Fill in the blanks
\frac3}h}3 = 4x + 2h - ___
Why: The quadratic part behaves exactly as in the previous example and contributes 4xh plus 2h squared. The linear part gives minus 3x minus 3h and then plus 3x, leaving minus 3h. Dividing every term by h gives 4x plus 2h minus 3.
\[ f(x+h) = 2x^2 + 4xh + 2h^2 - 3x - 3h \]
\[ f(x+h) - f(x) = 4xh + 2h^2 - 3h \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 80 — Exercise 168
Work out both blanks before checking. Everything you need is on the previous slide plus one new term.
Treat the two parts of the rule separately. The 2x squared part behaves exactly as before and contributes 4xh plus 2h squared. The minus 3x part gives minus 3 times x plus h, which is minus 3x minus 3h, and then subtracting the original minus 3x adds 3x back, leaving just minus 3h.
So the numerator is 4xh plus 2h squared minus 3h, and every term still carries a factor of h. Dividing gives 4x plus 2h minus 3. Notice the pattern: a linear term in the rule contributes its own coefficient, with no h attached, exactly as the straight-line example a few slides on will show.
Error analysis
Annotate
On: \( \frac{(x + h)^2 - x^2}{h} = \frac{x^2 + h^2 - x^2}{h} = \frac{h^2}{h} = h \)
Read the three lines and try to find the mistake before revealing the annotations. The algebra is short enough that the error has nowhere to hide, and yet it is the mistake most often made in this part of the course.
The square of a sum is not the sum of the squares. Expanding x plus h, all squared, gives three terms, and the middle term 2xh is the one that carries the whole answer. Drop it and you are left with h, which is wrong in a particularly nasty way: it still looks like a difference quotient, it still contains h, and nothing about its shape looks suspicious.
That is why the numerical test in the third annotation matters so much. Put x equal to 1 and h equal to 1: the true rate from 1 to 2 on the squaring rule is 3, the correct formula 2x plus h gives 3, and the wrong answer gives 1. Whenever you derive a formula, spend ten seconds testing it on numbers you can check by hand.
Worked example
Find the average rate of change on the interval from 9 to 9 plus h.
\[ a(t) = \frac{1}{t + 4} \]
Write both outputs
Why: At t equal to 9 the denominator is 13; at 9 plus h it is 13 plus h.
\[ a(9 + h) = \frac{1}{13 + h}, \qquad a(9) = \frac{1}{13} \]
Combine the numerator over a common denominator
Why: The common denominator is 13 times 13 plus h.
\[ \frac{1}{13 + h} - \frac{1}{13} = \frac{13 - (13 + h)}{13(13 + h)} \]
Simplify the top
Why: The two thirteens cancel, leaving minus h.
\[ \frac{-h}{13(13 + h)} \]
Divide by h, which multiplies by one over h
Why: The h on top cancels against the h underneath.
\[ \frac{-h}{13(13 + h)} \cdot \frac{1}{h} = \frac{-1}{169 + 13h} \]
Figure (svg): The curve one over t plus four, falling from 0.25 at t equals 0 to about 0.06 at t equals 12, with two close points at t equals 9 and t equals 11 joined by a nearly horizontal straight line.
Check the sign and a value
Why: The rule is decreasing for t bigger than minus 4, so every secant must fall and the answer must be negative, which it is. At h equal to 0.5 the formula gives minus 0.005698, and computing the quotient directly gives the same.
\[ \frac{-1}{169 + 6.5} = -0.005698 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 80 — Exercise 164
This is the type that looks hardest and is really just fraction arithmetic. The rule is a reciprocal, so the numerator of the quotient is a difference of two fractions, and the first job is to combine them over a common denominator.
That common denominator is 13 times the quantity 13 plus h. Once you put both fractions over it, the numerator becomes 13 minus the quantity 13 plus h, and the thirteens cancel, leaving minus h. Now the whole expression is minus h over something, all divided by h, and dividing by h is multiplying by one over h, so the h cancels and you are left with minus 1 over 169 plus 13h.
Check the sign, because it is free. The rule is falling everywhere to the right of minus 4, so every secant on it must slope downward and the answer must be negative for small positive h. Look at the figure: out at t equal to 9 the curve has almost flattened, so the secant is nearly horizontal and the answer is a small negative number, about minus five thousandths.
Prediction
\[ p(x) = 3x + 4, \qquad \text{on the interval from } 2 \text{ to } 2 + h \]
Predict first
What does the difference quotient simplify to?
Correct: 3, with no h left
Why: A straight line has the same steepness everywhere, so every secant on it IS the line, whatever the interval. The numerator is 3 times 2 plus h plus 4, minus 10, which is 3h, and dividing by h leaves 3. Any answer still containing h would be claiming that a line's steepness depends on how far you look.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 80 — Exercise 160
Predict before revealing. This one tests whether you understand what the difference quotient is measuring, rather than whether you can expand brackets.
A straight line has the same steepness everywhere. So any secant line drawn between two points of it is the line itself, and its slope is the line's slope whatever interval you chose and however wide it was. The answer must therefore be 3, with no h anywhere in it.
If you work it out algebraically the mechanism is visible: the numerator comes out as exactly 3h, and dividing by h leaves 3. Any answer still containing h would be claiming that a straight line is steeper over long intervals than over short ones, which is exactly what straight means it is not.
Tweak it
Parameter explorer
The curve plotted is the slope of the secant to the squaring rule joining the fixed input a to a second input x. Slide a, then read the height of the curve directly above the value x equals a. What do you notice?
\[ \frac{x^2 - ({a})^2}{x - {a}} = x + {a} \]
The curve on screen is not the parabola. It is the slope of the secant joining the fixed input a to a second input x, plotted as a function of that second input, and it simplifies to x plus a.
Set a to 1 and read off some heights. Above x equal to 3 the height is 4, above x equal to 2 it is 3, above x equal to 1.5 it is 2.5: those are exactly the three secant slopes in the earlier figure. Now slide your eye leftward towards x equal to a itself. The height there is 2, twice the fixed input.
Change a and the same thing happens: directly above the fixed input the height is always twice it. That number is the slope the secant pivots towards as the second point slides in, and in Calculus I you will learn to call it the derivative of the squaring rule. You have just read it off a graph with no limits and no new machinery.
Matching
Match the pairs
Why: A linear rule loses h entirely and returns its own slope. A quadratic returns something linear in x and h, and multiplying the rule by 3 multiplies the quotient by 3. A reciprocal needs a common denominator and comes out negative, because the reciprocal rule falls on each of its branches.
Match all four before checking. Two shortcuts make this quick. First, a linear rule always loses h completely and returns its own slope, so the constant answer belongs to the linear rule. Second, multiplying a rule by a constant multiplies its difference quotient by the same constant.
That second shortcut settles the two quadratics at once. The squaring rule gives 2x plus h, so three times the squaring rule gives three times that, which is 6x plus 3h. You do not have to expand anything.
The reciprocal is the odd one out and the only negative answer, which is the giveaway: the reciprocal rule falls on each of its branches, so every secant slope on it is negative. If you want to see where that answer comes from, it is the same common-denominator calculation as the previous worked example, with 4 replaced by 0.
Section
Part 3
Concept
Walk along the graph from left to right, that is, in the direction of increasing input. Where the walk goes uphill the function is increasing; where it goes downhill the function is decreasing; where it is level the function is constant.
Figure (svg): The curve x cubed minus twelve x on the window from minus 4 to 4. The stretch left of minus 2 and the stretch right of 2 are drawn in green and climb; the stretch between them is drawn in red and falls. The turning points at minus 2, height 16, and at 2, height minus 16, are marked.
Notice the link with the first part of this lesson: over any interval inside a rising stretch the average rate of change is positive, and over any interval inside a falling stretch it is negative.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 71-72 — Figure 1.45
The definitions are simpler than the notation makes them look. Put your finger on the left end of the curve and trace it rightward: where your finger rises the function is increasing, where it falls the function is decreasing, and where it stays level the function is constant.
The colours in the figure do that walk for you. The green stretches climb, the red stretch falls, and the two marked points are exactly where the direction changes. Those two points belong to neither colour, and that is not a drawing decision; it is the definition.
Connect this back to Part 1, because the two halves of the lesson are the same idea. On any interval inside a green stretch the average rate of change comes out positive; on any interval inside the red stretch it comes out negative. Increasing and decreasing are just the local version of the sign of the rate.
Notation
Annotate
On: \( b > a \;\Longrightarrow\; f(b) > f(a) \)
Read it as: whenever one input from the interval is bigger than another, the output at the bigger input is higher. An interval is always named with inputs, never with outputs, and that is the single most common slip in this part of the section.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 73 — Local Minima and Local Maxima box
Read the symbols slowly the first time: for any two inputs in the interval, if the second is bigger than the first, then the output at the second is bigger than the output at the first. That is all the notation is saying.
The words for all are doing heavy lifting. It is not enough that the function happens to be higher at the right-hand end than at the left; every pair of inputs inside the interval must behave. That is precisely why a positive average rate of change is not the same as increasing, which you will see broken on purpose two slides from now.
Two habits to lock in. Intervals of increase and decrease are always written with inputs, never with outputs. And they are always written with round brackets, because the endpoints, being the turning points, are neither increasing nor decreasing.
Worked example
The graph shows a function p. Identify the intervals on which the function appears to be increasing.
Figure (svg): A curve on the window from 0 to 5. It falls to a low point at t equals 1, climbs to a high point at t equals 3, dips to a second low point at t equals 4, then climbs steeply. The three turning points are marked.
Find every place the direction changes
Why: Three turning points, at the inputs 1, 3 and 4.
\[ t = 1, \quad t = 3, \quad t = 4 \]
Describe each stretch between them
Why: Read the colours: downhill, uphill, downhill, uphill.
\[ \text{down on } (0,1), \; \text{up on } (1,3), \; \text{down on } (3,4), \; \text{up on } (4, \infty) \]
Collect the increasing stretches
Why: Two of the four.
\[ \text{increasing on } (1,3) \; \text{and} \; (4, \infty) \]
Check the bracket type
Why: Round brackets throughout: at t equal to 1, 3 and 4 the function is neither increasing nor decreasing, so those inputs belong to no interval. The function is constant nowhere, so no stretch is left over.
\[ (1,3) \text{ and } (4,\infty) \text{ are open; } t = 1, 3, 4 \text{ are excluded} \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 73 — Example 1.35 and Figure 1.48
Work along the curve from left to right and mark every place the direction changes. There are three of them, at the inputs 1, 3 and 4, and they cut the picture into four stretches that alternate: down, up, down, up.
Now just collect the up ones. The function is increasing from 1 to 3, and then again from 4 onward, and in interval notation that is the interval from 1 to 3 together with the interval from 4 to infinity. The word together is important: this is two separate intervals, not one, and it is wrong to write a single interval from 1 to infinity.
Look hard at the brackets. They are round at both ends, everywhere. At the inputs 1, 3 and 4 the graph is turning, so at those three inputs the function is neither increasing nor decreasing and they belong to no interval at all. Infinity always gets a round bracket too, since it is not a number the function can reach.
Trap
Asked where the graph of Example 1.35 is increasing, a student writes:
\[ \text{increasing on } (-1.3,\; 1.37) \]
Wrong. Those are the two heights, not the two inputs.
An interval of increase lives on the horizontal axis. It is a set of inputs, and it is read off the horizontal axis, never the vertical one.
\[ \text{increasing on } (1, 3) \]
The mirror-image slip is to report an extremum as an input. Intervals use inputs; extreme values are outputs.
This one is not an arithmetic slip; it is a confusion about what the question is asking, and it is worth naming because once you see it you stop making it.
An interval of increase is a set of inputs. It lives on the horizontal axis, and you find it by reading along that axis. The heights, minus 1.3 and 1.37, live on the vertical axis and answer an entirely different question.
The mirror-image slip is coming in Part 4, where people report a local maximum as an input instead of an output. Hold the two rules together and neither can catch you: intervals are made of inputs, extreme values are outputs.
Sorting
Sort into buckets
Using the graph of Example 1.35, with turning points at 1, 3 and 4, sort each interval or input.
Sort all six before checking, and keep the picture from Example 1.35 in front of you: down to the input 1, up to 3, down to 4, up after that.
Four of the items are intervals and they sort themselves once you know the four stretches. The other two items, the single inputs 1 and 3, both go in the third bucket. A single input is not an interval at all, and even if it were, at a turning point the function is changing from one behaviour to the other, so neither word applies.
If you put either single input in the increasing or decreasing bucket, go back to the formal definition and see what it needs: two different inputs in an interval, with one bigger than the other. One point cannot supply that.
Worked example
Graph the function below, estimate its local extrema, and use them to give the intervals of increase and decrease.
\[ f(x) = x^3 - 6x^2 - 15x + 20 \]
Figure (svg): The curve x cubed minus six x squared minus fifteen x plus twenty, rising to a high point at x equals minus 1 of height 28, falling to a low point at x equals 5 of height minus 80, then rising again.
Read the turning points off the graph
Why: A high point on the left and a low point on the right.
\[ (-1,\; 28) \quad\text{and}\quad (5,\; -80) \]
Split the number line at those two inputs
Why: Three stretches, and the graph's direction alternates.
\[ (-\infty,-1), \quad (-1,5), \quad (5,\infty) \]
Name the behaviour on each
Why: Up, down, up, reading left to right.
\[ \text{increasing on } (-\infty,-1) \text{ and } (5,\infty); \quad \text{decreasing on } (-1,5) \]
Check the two heights by substituting
Why: Both turning points have whole-number coordinates here, so the graph reading can be confirmed exactly.
\[ f(-1) = -1 - 6 + 15 + 20 = 28, \qquad f(5) = 125 - 150 - 75 + 20 = -80 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 75 — Try It 1.24
This exercise assumes a graphing tool, and using one is the intended method, not a shortcut. Type the rule in, look at the shape, and read off the two turning points.
The graph gives a high point near the input minus 1 and a low point near the input 5. Those two inputs cut the whole number line into three stretches, and because the graph of a cubic with two turning points must alternate, the behaviour is up, down, up. So it increases before minus 1, decreases between minus 1 and 5, and increases after 5.
The check is available here because the numbers happen to be whole. Substituting minus 1 gives 28 and substituting 5 gives minus 80, exactly the heights the graph shows. Most cubics are not so obliging, and then the graph reading is all you have at this level; finding turning points exactly is a calculus job.
Counterexample
Figure (svg): The parabola y equals x squared on the window from minus 1.6 to 2.4, with points marked at minus 1, height 1, and at 2, height 4. The straight line joining them climbs, but the curve falls to the origin before it climbs.
Discussion prompt
Someone claims: if the average rate of change of f over the interval from a to b is positive, then f is increasing on that interval. Use the squaring rule on the interval from minus 1 to 2 to break the claim, then say in one sentence what the average rate of change really guarantees.
Write your counterexample before revealing, and choose something simple: the squaring rule on an interval straddling the origin does it in one line.
From the input minus 1 to the input 2 the output goes from 1 up to 4, so the average rate of change is 1, comfortably positive. And yet, as the red part of the curve shows, the function falls throughout the first third of that interval before it climbs. Positive average, not increasing.
The reason is exactly the for all in the formal definition. An average rate of change compares only the two endpoints; increasing demands that every later output beat every earlier one. What a positive average rate really guarantees is one narrow thing: the output ended higher than it started. Nothing about the route.
Prediction
Predict first
At an input where a graph changes from rising to falling, what is the function doing?
Correct: Neither increasing nor decreasing there
Why: Increasing and decreasing are properties of intervals, not of single inputs, and the definitions need an open interval on which every later output beats every earlier one. No open interval around a turning point does that. This is exactly why the intervals of increase are written with round brackets.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 72 — the paragraph after Figure 1.45
Commit to an answer first. The wording of the options is deliberately close, and the one that tempts people is the third: it was increasing just before, so surely it is still increasing at the turning point itself.
Go back to the definition to settle it. Increasing is a property of an interval, and it requires an open interval on which every later output beats every earlier one. Around a turning point no such interval exists, however small you make it, because on one side the outputs are going the other way. So the answer is neither.
This is not pedantry; it is why every interval of increase in this section is written with round brackets. The turning points are excluded from the intervals on either side of them, and they get their own name instead: local extrema, which is where the next part goes.
Section
Part 4
Concept
An input where the function changes from increasing to decreasing gives a local maximum; the other way round gives a local minimum. Together they are called the local extrema, and the singular is extremum.
Figure (svg): A curve with a hump. The top of the hump is marked, a dashed horizontal line runs through it, and two vertical dashed lines at a and c bracket an open interval containing the hump. Inside that interval no part of the curve rises above the dashed horizontal line.
local maximum — The function f has a local maximum at the input b if there is an open interval containing b on which no output beats f(b). The local maximum itself is the output f(b), not the input b.
\[ \text{local max at } b: \quad f(x) \le f(b) \text{ for all } x \text{ in some } (a,c) \text{ with } a < b < c \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 72-73 — Figures 1.46, 1.47 and the definition box
The definition has three moving parts and the figure shows all three. There is the point itself, at the top of the hump. There is an open interval around it, marked by the two vertical dashed lines. And there is the claim: inside that interval, nothing on the curve gets above the dashed horizontal line through the point.
Read the last annotation of the definition carefully, because it is where marks are lost. The local maximum is the output value, the height of the point. The input is where it occurs. A complete answer always says both: the local maximum is such-and-such, and it occurs at this input.
And notice what the figure allows. Further to the right the curve climbs well above the dashed line, and the point at the top of the hump is still a local maximum. It only ever had to beat its own neighbourhood.
Intuition
The word local is doing real work. Like the summit of one hill on a long walk, a local maximum only has to beat the ground immediately either side of it. Two miles further on there may be a much higher hill, and the first summit is still a summit.
That is why the definition demands an open interval containing the input, and nothing more. Shrink the interval enough and the point wins; that is the whole test. Reading the graph, the question to ask at a candidate point is simply whether the curve is lower on both sides nearby.
It is also why an endpoint of a window is never called a local extremum in this book: there is no open interval around it that stays inside the domain, so there is no neighbourhood on both sides to beat.
The hill-walking picture is worth keeping. A local maximum is the summit of one hill: higher than the ground immediately either side, and completely indifferent to what the mountains two valleys over are doing.
That is why the definition demands only that some open interval exists on which the point wins. You get to choose the interval, and you may choose it as small as you like. So the practical test when reading a graph is simply this: is the curve lower on both sides, close in? If yes, local maximum, whatever happens further away.
It also explains a convention that otherwise looks arbitrary. This book does not call an endpoint of a domain a local extremum, because there is no open interval around an endpoint that stays inside the domain: there is no neighbourhood on both sides for it to beat. Endpoints get their turn in the next part, where they compete for the absolute extrema instead.
Worked example
For the function f whose graph is shown, find all local maxima and minima.
Figure (svg): The curve three x minus x cubed, coming down from the upper left, bottoming out at the point minus 1, minus 2, rising to the point 1, 2, then falling away to the lower right. Both turning points are marked and labelled.
Find every input where the direction changes
Why: Two of them, one on each side of the origin.
\[ x = -1 \quad\text{and}\quad x = 1 \]
Classify the one at minus 1
Why: The graph falls into it and rises out of it, so it is the lowest point of a small interval around minus 1.
\[ \text{local minimum at } x = -1, \text{ value } f(-1) = -2 \]
Classify the one at 1
Why: The graph rises into it and falls out of it.
\[ \text{local maximum at } x = 1, \text{ value } f(1) = 2 \]
Check that local really is only local
Why: Far to the left the curve is well above the height 2, and far to the right it is well below minus 2. Neither extremum is the best or worst overall, and neither has to be: the test is only against nearby points.
\[ f(-2.4) = 6.62 > 2, \qquad f(2.4) = -6.62 < -2 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 75 — Example 1.37 and Figure 1.51
Find the direction changes first. There are two, one on each side of the origin, and they are the only candidates.
Classify each by how the curve arrives and leaves. At the input minus 1 the curve falls in and rises out, so that is the bottom of a valley: a local minimum, of value minus 2. At the input 1 the curve rises in and falls out, so that is the top of a hill: a local maximum, of value 2. Say both halves of each sentence, the value and the location.
Then look at the far ends of the curve, because they make the point of the word local. Out at the left the curve is well above the height 2, and out at the right it is well below minus 2. So the local maximum is not the largest output the function ever produces, and the local minimum is not the smallest. Neither has to be.
Error analysis
Annotate
On: \( \text{local max} = 1, \qquad \text{local min} = -1 \)
Read the claim before opening the annotations and decide what is wrong with it. The two numbers quoted are both correct numbers from this problem, which is what makes the error so easy to make and so easy to miss.
The question asked for the local maxima and minima, and those words name outputs: heights, read off the vertical axis. What has been given instead are the inputs, read off the horizontal axis. Those are the locations of the extrema, not the extrema.
The safe habit is to write the full sentence every time: the local maximum is 2, and it occurs at the input 1. Two pieces of information, clearly labelled, and no marker can dock you for ambiguity. Pair that with the rule from Part 3 and you are covered both ways: intervals take inputs, extreme values are outputs.
Worked example
Graph the function below, then use the graph to estimate its local extrema and the intervals on which it is increasing.
\[ f(x) = \frac{2}{x} + \frac{x}{3} \]
Figure (svg): Two separate branches of the curve two over x plus x over three. The right branch dips to a low point near x equals 2.449 at height 1.633; the left branch rises to a high point near x equals minus 2.449 at height minus 1.633.
Describe what the graph shows
Why: Two separate branches, with a low point on the right branch and a high point on the left.
\[ \text{low point near } x \approx 2.4, \qquad \text{high point near } x \approx -2.4 \]
Read the estimates off a graphing tool
Why: Two different tools agree only to about four decimal places, because each uses its own search algorithm.
\[ \text{local min } \approx 1.633 \text{ at } x \approx 2.449; \quad \text{local max } \approx -1.633 \text{ at } x \approx -2.449 \]
Name the increasing intervals
Why: Outside the two turning points the curve climbs; between them, on each branch, it falls.
\[ \text{increasing on } (-\infty,-2.449) \text{ and } (2.449, \infty) \]
Check the estimate against the exact value
Why: The exact location is the square root of 6, which calculus supplies and precalculus cannot. Its decimal is 2.4494897, matching the tool's 2.449, and the height there is 1.6329932, matching 1.633.
\[ \sqrt{6} = 2.4494897\ldots, \qquad f\left(\sqrt{6}\right) = 1.6329932\ldots \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 74-75 — Example 1.36 and Figures 1.49, 1.50
Two things make this example different. First, the function has two separate branches, because the input 0 is not allowed; the turning points sit one on each branch. Second, the turning points are not at nice inputs, so a graphing tool is not a convenience here, it is the method.
The book's own screenshots make a point worth noticing: two different calculators agree only to about four decimal places, because each hunts for the low point with its own search algorithm. So the honest way to report these is as estimates, about 1.633 at about 2.449, and to say that they are estimates.
The exact location is the square root of 6, which the book mentions and cannot prove, because finding it requires calculus. When you meet this function again in Calculus I you will get 2.4494897 in two lines of algebra. For now, notice that the decimal matches the tool's answer, which is the best check available at this level.
Discrimination
Sort into buckets
Each description is of a single point on a graph. Does the function have a local extremum there?
Sort all six before checking. Two of them, the first and the third, are just the definitions of local maximum and local minimum in words, so they go in the yes bucket immediately. The sixth is the definition of a local minimum again, phrased in terms of neighbours instead of direction, so it belongs there too.
The second item is the interesting one, and there is a worked example of it on the next slide. A graph can flatten out momentarily and then carry on in the same direction. Flat is not the same as turning: if the curve was going downhill before and is going downhill after, it was never the lowest point nearby.
The fifth item is a different confusion. Crossing the horizontal axis means the output is 0 there, which is a statement about the height, not about the direction of travel. A function can cross zero while climbing steadily and have no extremum anywhere near. And the fourth, the endpoint, is excluded by this book's convention, as the earlier slide explained.
Worked example
Use a graphing utility to estimate the local extrema and the intervals of increase and decrease.
\[ f(x) = x^4 - 4x^3 + 5 \]
Figure (svg): The curve x to the fourth minus four x cubed plus five. It falls from the left, flattens briefly at x equals 0 without turning, keeps falling to a low point at x equals 3 of height minus 22, then rises steeply.
Find the flat places on the graph
Why: Two of them, at the inputs 0 and 3.
\[ x = 0 \quad\text{and}\quad x = 3 \]
Test the one at 0
Why: The curve is falling just left of 0 and still falling just right of it, so no direction change happens.
\[ f(-0.5) = 5.56 > f(0) = 5 > f(0.5) = 4.56 \]
Test the one at 3
Why: The curve falls in and climbs out, so this one is a genuine local minimum.
\[ f(2.5) = -18.44 > f(3) = -22 < f(3.5) = -16.06 \]
Name the intervals
Why: Decreasing all the way to 3, then increasing; the flat spot at 0 does not split anything.
\[ \text{decreasing on } (-\infty, 3); \quad \text{increasing on } (3, \infty) \]
Check the local minimum by substituting
Why: Eighty-one minus one hundred and eight plus five is minus twenty-two, matching the graph, and both neighbours are higher.
\[ f(3) = 81 - 108 + 5 = -22 \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 82 — Exercise 188
Graph it first and you will see two places where the curve goes flat, at the inputs 0 and 3. The whole exercise is deciding which of them is a genuine local extremum.
The test is to look just either side. Near 0 the outputs are 5.56, then 5, then 4.56: falling, flat, still falling. The curve levelled off and then carried straight on downhill, so that point beats nothing and is no extremum. Near 3 the outputs are minus 18.44, then minus 22, then minus 16.06: falling, then rising, so that one really is a local minimum.
Because the flat spot at 0 is not a direction change, it does not split the intervals either. The function decreases all the way up to the input 3 and increases after it: two intervals, not four. This is the example to remember whenever you are tempted to equate flat with turning.
Edge cases
Parameter explorer
This family of cubics is x cubed minus three c x. Slide c from 0 upward and watch the middle of the curve. For which values of c does the graph have a local maximum and a local minimum, and for which does it have neither?
\[ y = x^3 - 3({c})x \]
Start with the slider at 1 and look at the shape: a rising curve, a hump, a dip, then rising again, with two turning points either side of the origin. Now drag the parameter down slowly.
As the parameter shrinks towards 0 the hump and the dip move towards each other and flatten out. At exactly 0 they have merged into a single flat spot at the origin, and the curve is climbing on both sides of it: that is precisely the flat-but-not-turning situation from the previous slide, in a whole family of curves at once. Push the parameter negative and even the flat spot disappears; the curve climbs everywhere.
Push it the other way instead and the two turning points slide apart. Their inputs are the positive and negative square roots of the parameter, so at 1 they sit at plus and minus 1, and at 4 they would sit at plus and minus 2. One picture, the whole family, and the moment of transition visible as it happens.
Section
Part 5
Concept
A local extremum beats its neighbours. An absolute extremum beats every other output the function has, over the whole domain. As before, the extremum is the output value, and it occurs at one or more inputs.
\[ \text{absolute max at } c: \quad f(c) \ge f(x) \text{ for every } x \]
\[ \text{absolute min at } d: \quad f(d) \le f(x) \text{ for every } x \]
Figure (svg): The curve x cubed minus six x squared plus nine x plus one drawn only from x equals 0 to x equals 5, with filled dots at both ends. It rises to a hump of height 5 at x equals 1, dips to height 1 at x equals 3, then climbs to height 21 at the right end.
Note what the picture shows: the graph's local maximum, the hump at the input 1, is not the absolute maximum. The tallest point of the whole graph is the right-hand endpoint, where the curve is not turning at all.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 77 — Absolute Maxima and Minima box, Figure 1.55
The wording changes in exactly one place. A local maximum beats every output on some open interval around it; an absolute maximum beats every output the function has anywhere in its domain. Everything else, including the rule that the extremum is the output and not the input, is unchanged.
The figure is built to make one point. The hump at the input 1 is a genuine local maximum, height 5, and it is nowhere near being the absolute maximum, because the right-hand endpoint of the domain reaches 21. The tallest point of the whole graph is a place where the curve is not turning at all.
That is the structural lesson of this part. Local extrema live at turning points; absolute extrema live at turning points or at endpoints, and you will not know which until you have compared all the heights.
Intuition
Local extrema can be found by looking at the graph one bump at a time. Absolute extrema cannot: they are a competition between every output the function has, so the answer depends on which inputs are allowed.
Shrink the domain to a closed window and every candidate is either a turning point inside the window or one of the two endpoints, so the competition is short. Open the domain up and there may be no winner at all: the cubing rule climbs forever in both directions and has neither an absolute maximum nor an absolute minimum.
So the procedure has two halves. Collect the turning points, then collect the endpoints, then compare the heights. Skipping the second half is how most absolute extrema get lost.
Local extrema can be found one bump at a time, which makes them comfortable. Absolute extrema cannot, because the question is a competition between every output the function has, and which outputs exist depends entirely on which inputs are allowed.
Close the domain into a window with two endpoints and the competition becomes short and reliable: the winner is either a turning point inside the window or one of the two ends. That is a finite list, and comparing it is arithmetic.
Leave the domain open and there may be no winner at all. The cubing rule runs off to infinity upward and downward, so it has neither an absolute maximum nor an absolute minimum, and the honest answer to a question about them is that none exists. Saying none exists is a complete answer, not a failure to find one.
Worked example
For the function f whose graph is shown, find all absolute maxima and minima.
Figure (svg): A curve on the window from minus 2.6 to 3 with two equal peaks of height 16, at x equals minus 2 and x equals 2, a valley of height 1 at the origin, and a steep plunge to a filled endpoint at 3 of height minus 10.
List every candidate height
Why: Two peaks, one valley, and the two ends of the drawn domain.
\[ f(-2) = 16, \; f(0) = 1, \; f(2) = 16, \; f(3) = -10 \]
Pick the largest
Why: The two peaks tie at the same height.
\[ \text{absolute maximum } 16, \text{ at } x = -2 \text{ and } x = 2 \]
Pick the smallest
Why: The right-hand endpoint plunges below everything else.
\[ \text{absolute minimum } -10, \text{ at } x = 3 \]
Check that nothing else competes
Why: The valley at the origin, height 1, is a local minimum but not the absolute one, since minus 10 is lower. Notice that the absolute minimum here sits at an endpoint, not at a turning point, so a search that looked only at the turning points would have returned the wrong answer.
\[ 1 > -10 \;\Longrightarrow\; \text{the valley is only local} \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 78 — Example 1.38 and Figure 1.56
Build the candidate list first, before deciding anything: the two peaks, the valley between them, and both ends of the drawn domain. Five heights. Then pick the largest and the smallest.
The two peaks tie at exactly 16, and the right way to report that is that the absolute maximum is 16 and it occurs at two inputs, minus 2 and 2. One value, two locations. Ties are perfectly ordinary here, because the definition asks the output to beat or equal every other output, not to beat it strictly.
The absolute minimum is the interesting half. It is minus 10, and it sits at the right-hand endpoint, where the graph is plunging and not turning at all. A search that looked only at the turning points would have reported the valley at height 1 and been wrong by eleven units. Endpoints always join the list.
Comparison
Comparison matrix
| question | local extremum | absolute extremum |
|---|---|---|
| what it must beat | nearby outputs only | every output |
| where you look | turning points | turning points and endpoints |
| is it an input or an output? | output | output |
| can there be several? | yes, one per turning point | yes, the same value at several inputs |
| must one exist? | no | no |
Fill the blanks before checking, and if a row makes you hesitate, go back to the definition rather than to your memory of the last example.
The second row is the one that matters most in practice. Local extrema are found at turning points, full stop. Absolute extrema are found at turning points and at endpoints, and forgetting the second half is the commonest error in this part of the section.
The last row surprises people. Neither kind has to exist. A straight line has no local extrema at all, and the cubing rule has no absolute ones. When the honest answer is none, write none.
Trap
Asked for the absolute extrema of the cubic below on the window from 0 to 6, one turning point inside it gets found, and the search stops:
\[ f(x) = x^3 - 6x^2 - 15x + 20 \]
\[ \text{both extrema at } x = 5? \]
Wrong. The only interior turning point is the low one; the high one, at minus 1, is outside the window.
On a closed window the endpoints are candidates too, and here the left endpoint wins the maximum outright.
\[ f(0) = 20 \]
\[ f(5) = -80, \quad f(6) = -70 \]
Absolute maximum 20 at the input 0; absolute minimum minus 80 at the input 5. Always add the endpoints to the list before comparing.
This is the trap the whole of Part 5 has been building towards, and it is not carelessness: the hard part was done correctly, and then a slightly different question got answered.
On the window from 0 to 6 there is exactly one turning point, the low one at the input 5. The high one, at minus 1, is outside the window and is simply unavailable. So the list of candidates cannot possibly be complete with only one entry on it; a maximum and a minimum both have to come from somewhere.
Add the two endpoint heights, 20 at the input 0 and minus 70 at the input 6, and the answer falls out: the absolute maximum is 20, at the left endpoint, and the absolute minimum is minus 80, at the turning point. Make evaluating the endpoints a reflex the moment you see a closed window.
Worked example
Take the cubic from Try It 1.24 and restrict its domain to the closed interval from 0 to 6. Find the absolute extrema.
\[ f(x) = x^3 - 6x^2 - 15x + 20, \qquad 0 \le x \le 6 \]
Figure (svg): The cubic x cubed minus six x squared minus fifteen x plus twenty drawn only from x equals 0 to x equals 6, with filled dots at both ends. It falls from height 20 at the left end to a low point of minus 80 at x equals 5, then rises a little to minus 70 at the right end.
List the turning points inside the window
Why: Of the two turning points, only the low one at 5 lies between 0 and 6.
\[ x = 5: \; f(5) = 125 - 150 - 75 + 20 = -80 \]
Add the two endpoints
Why: Substitute 0 and 6.
\[ f(0) = 20, \qquad f(6) = 216 - 216 - 90 + 20 = -70 \]
Compare the three heights
Why: Twenty is the largest, minus eighty the smallest.
\[ -80 < -70 < 20 \]
State the answer with both parts
Why: Value first, then where it happens.
\[ \text{absolute max } 20 \text{ at } x = 0; \quad \text{absolute min } -80 \text{ at } x = 5 \]
Check against the unrestricted graph
Why: Without the window this cubic has no absolute extrema at all, since it runs off to infinity in both directions; its local maximum 28 at minus 1 is outside the window and plays no part here. Closing the window is what created the answer.
\[ f(-1) = 28 \text{ is unavailable: } -1 \text{ is not in } [0,6] \]
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, pp. 77-78 — the Absolute Maxima and Minima box, applied to Try It 1.24
This is the same cubic you graphed in Try It 1.24, with one change: the domain has been cut down to the inputs from 0 to 6. Watch how much that one change alters the answers.
Work the list. Inside the window there is one turning point, the low one at the input 5, of height minus 80. The two endpoints give 20 and minus 70. Compare the three and the largest is 20, at the left endpoint, while the smallest is minus 80, at the turning point.
The last step is the one worth remembering. Without the window this cubic has no absolute extrema at all, because it runs off to infinity in both directions, and its local maximum of 28 at the input minus 1 is not even available here. The window did not reveal the answer; it created it. Whenever a question gives you a closed interval, that interval is part of the problem, not decoration.
Prediction
Predict first
Taking each rule on its whole natural domain, which one has an absolute maximum?
Correct: The negative of the squaring rule
Why: Turning the parabola upside down puts its vertex at the top, so the output 0 at the input 0 beats every other output: that is an absolute maximum. The squaring rule has an absolute minimum instead, 0 at the input 0. The cubing rule runs to infinity both ways and has neither, which is the example the book names. The reciprocal rule grows without bound near 0 on each branch, so it has neither.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 77 — the remark that the cubing rule has no absolute extremum
Choose before revealing, and sketch each of the four in your head rather than reasoning about them in words. Three of the four are toolkit functions you have already drawn many times.
An upside-down parabola has its vertex at the top, and every other output is below it, so the output 0 at the input 0 is an absolute maximum. The ordinary squaring rule is the same picture flipped: it has an absolute minimum of 0 and no maximum, since it climbs without bound in both directions.
The cubing rule is the example the book names for having neither, because it runs to infinity upward and downward. The reciprocal rule also has neither: just to the right of 0 it grows without bound, and just to the left it falls without bound. Two of these four functions have an absolute extremum of one kind and none of the other, which is a good reminder that the two questions are separate.
Section
Part 6
Pattern
Figure (svg): A flow diagram in five boxes: pick two inputs and divide the change in output by the change in input; then walk left to right to name rising and falling stretches; then mark where the direction changes as local extrema; then add the endpoints of the window; then compare every candidate height to find the absolute extrema.
This is the order to work in whenever a question hands you a graph and asks you to describe it. Each step feeds the next, so doing them out of order costs time and usually costs an endpoint.
The middle three steps are the ones that get muddled. Name the intervals with inputs, in round brackets. Report each local extremum as an output value together with the input where it occurs. Then, and only then, add the endpoint heights to the list, which is a step that has nothing to do with the local extrema and is easy to skip because of that.
The final line of the figure is the sentence to memorise, because it is what most of the marks in this section are actually for: intervals take inputs, extreme values are outputs.
Ranking
Put in order
Order the steps of a complete description of a graph on a closed window.
Why: The direction changes come first because everything else is built on them: they cut the domain into intervals and they are the local extrema. The endpoints are evaluated next, because they are the candidates that the turning points alone would miss, and only when every candidate is on the list can the comparison for the absolute extrema be made.
Drag the five steps into order before checking. Only two orderings are genuinely tempting, and the difference between them is where the endpoints go.
Everything hangs off the direction changes, so they come first: they cut the domain into intervals and they are themselves the local extrema. Naming the intervals and reporting the local extrema are two readings of the same information, and either order would be defensible, but describing the intervals first matches the way you read the graph.
The endpoints must be evaluated before the final comparison and cannot be folded into the local extrema step, because an endpoint is not a local extremum. It is only a candidate for the absolute extrema, and the comparison is the last thing you do, once the list is complete.
Check
Check your understanding
What is the average rate of change of the function five minus two x squared on the interval from minus 2 to 4?
Answer: A
Why: The outputs are 5 minus 2 times 4, which is −3, at the input −2, and 5 minus 2 times 16, which is −27, at the input 4. The change in output is −27 minus −3, that is −24, over a change in input of 6, giving −4.
Both outputs need computing before anything else, and the negative input is where the arithmetic goes wrong. Two times minus 2 squared is 2 times 4, which is 8, so the output there is minus 3, not 13.
Then the run. It is 4 minus negative 2, which is 6, not 2. Two of the wrong options exist precisely to catch people who forgot to divide at all, or who divided by the wrong run.
A glance at the shape settles the sign without any work: this is an upside-down parabola, and on this interval it ends far lower than it started, so the rate has to be negative.
Check
Check your understanding
A function is increasing on the interval from 1 to 3 and decreasing on the interval from 3 to 6. Which statement must be true?
Answer: A
Why: Rising into an input and falling out of it is exactly the definition of a local maximum there: on some open interval around 3 no output beats the one at 3.
The question asks what must be true, so the job is to eliminate the options that could be true rather than the ones that are false. Only one of the four is forced by the information given.
Rising into an input and falling out of it is the definition of a local maximum there, so that one is forced. A local minimum would need the opposite order, so it is simply wrong rather than uncertain.
The other two are both could-be-true traps. Nothing at all is known about the function outside the stretch from 1 to 6, so it may climb higher elsewhere and the absolute maximum claim is unsupported. And the average rate over the whole stretch depends on how the two endpoint heights compare, which the question deliberately does not tell you.
Check
Check your understanding
On the closed window from 0 to 6 a function has a local maximum of 12 at the input 2, a local minimum of 1 at the input 4, and endpoint values f(0) = 5 and f(6) = 15. What is the absolute maximum?
Answer: A
Why: Every candidate is on the list: 5, 12, 1 and 15. The largest is 15, and it happens at the endpoint 6. An endpoint is not called a local extremum, but it competes for the absolute ones.
This question hands you the complete candidate list, which is the part you usually have to build yourself: 5 and 15 at the two endpoints, 12 and 1 at the two turning points. All that remains is to pick the largest.
The tempting wrong answer is 12, because it is the largest of the turning-point heights and turning points are what Part 4 trained you to look at. The endpoint at the input 6 is higher, and endpoints always join the comparison for absolute extrema even though they are never called local extrema.
One of the options offers the input 6 as the answer. That is the input-versus-output confusion again, in its purest form. The extremum is the height 15; the input 6 is where it happens.
Explain it to yourself
Discussion prompt
In two or three sentences, explain to someone who missed the lesson why two very different journeys can have the same average rate of change, and what extra information the increasing and decreasing intervals give that the average rate cannot.
Write your explanation out before revealing. Explaining an idea in your own words is a much harder test of understanding than answering a question about it, which is exactly why it is worth doing.
The key observation is that the formula contains exactly two function values, the ones at the two ends. Any two functions that agree at those two inputs produce the same number, no matter how wildly they differ in between, which is the picture from early in the lesson.
The second half of the question is the point of the whole lesson. The intervals of increase and decrease describe the route rather than the destination; the local and absolute extrema describe the highs and lows along the way. Those are exactly the pieces of information a single average rate cannot carry, which is why this section needs both halves.
Real world
Figure (svg): A decay curve starting at 100 milligrams and halving every ten days, drawn out to day 30. Two points at day 5 and day 15 are marked and joined by a straight line that falls.
Discussion prompt
The graph shows a radioactive sample decaying from 100 milligrams, halving every ten days. Estimate the average decay rate from day 5 to day 15, with units, and then say whether the same calculation from day 20 to day 30 would give a bigger or a smaller number.
OpenStax, Precalculus, §1.3 Rates of Change and Behavior of Graphs §1.3, p. 83 — Exercises 199 to 203, real-world applications
Read the two heights off the curve first, then do the division. At day 5 the sample is at about 70.7 milligrams and at day 15 it is at about 35.4, a fall of about 35 milligrams over ten days, so the average decay rate is about minus 3.5 milligrams per day.
Now the second half of the question, which is where the understanding is. The same calculation from day 20 to day 30 runs from about 25 milligrams down to about 12.5, a fall of only about 12.5 over the same ten days, so the rate is about minus 1.25 milligrams per day. Same length of interval, a rate less than half as steep.
That is decay in a nutshell, and it is why the curve flattens: what the sample loses in a fixed time is proportional to how much is left, so the later you look, the gentler the secant. The same arithmetic describes a cooling cup of coffee, a discharging capacitor and a drug clearing from the bloodstream.
Exit ticket
\[ f(x) = x^3 - 3x^2 + 4, \qquad \text{on the closed window } -1 \le x \le 3 \]
Discussion prompt
Find the average rate of change over the whole window, name the intervals of increase and decrease, give every local extremum, and give the absolute extrema.
This puts the whole lesson in one problem, so work through it in the order from the pattern slide rather than jumping to whichever part looks easiest.
The average rate over the whole window uses only the two endpoint heights, 0 at the input minus 1 and 4 at the input 3, giving a rate of 1. The turning points are at the inputs 0 and 2, with heights 4 and 0, so the function increases, decreases and increases again, and the local maximum is 4 at the input 0 while the local minimum is 0 at the input 2.
The absolute extrema need the endpoint heights added to the list, and here something neat happens: the endpoint heights are 0 and 4, exactly matching the two turning-point heights. So the absolute maximum is 4, reached at both the input 0 and the input 3, and the absolute minimum is 0, reached at both the input minus 1 and the input 2. Ties in both directions, and a good reminder that an extreme value can occur at more than one input.
Recap
| idea | what it is | what to report |
|---|---|---|
| average rate of change | change in output over change in input, the slope of a secant line | one number, with units, and the interval it belongs to |
| difference quotient | the same slope with the second input written as a plus h | an expression in a and h, after the h has cancelled |
| increasing / decreasing | the direction of travel as the input grows | open intervals of INPUTS |
| local extremum | an output that beats its neighbours on some open interval | the OUTPUT value, and the input where it occurs |
| absolute extremum | an output that beats every other output in the domain | the OUTPUT value; check the endpoints too |
\[ \frac{\Delta y}{\Delta x} = \frac{f(x_2) - f(x_1)}{x_2 - x_1} \qquad \text{and} \qquad \frac{f(a+h) - f(a)}{h} \]
Next, Section 1.4 builds new functions out of old ones, and Chapter 12 takes the difference quotient to its limit as the gap closes, which turns the secant into a tangent and this lesson into calculus.
OpenStax Algebra and Trigonometry 2e, §3.3 Rates of Change and Behavior of Graphs §3.3 — the same material in Algebra and Trigonometry
Two computations and three descriptions. The average rate of change is the slope of the secant line joining two points of the graph; the difference quotient is that same slope with the gap written as a letter, which is what makes it a formula and what makes calculus possible.
The three descriptions are where the marks are. Intervals of increase and decrease are open intervals of inputs. Local extrema are output values, found at turning points, and they only have to beat their neighbours. Absolute extrema are output values too, found at turning points and endpoints, and they have to beat everything.
If you carry one sentence out of this lesson, make it this one: intervals take inputs, extreme values are outputs. Almost every wrong answer in this section is one of those two the wrong way round. Section 1.4 goes on to build new functions from old ones, and Chapter 12 closes the gap in the difference quotient and turns the secant into a tangent.
Want this taught 1-on-1? Alexander tutors Precalculus — $55/session, free consultation.